Research And Application Of Computing Offloading And Resource Allocation Based On Mobile Edge Computing

In the future,new computing-intensive applications in mobile networks require stronger computing power and lower latency.However,due to the high battery loss and insufficient computing power of mobile users’ devices,users cann ot enjoy higher quality service experience.Mobile edge computing technology can effectively solve the problems of computing performance,resource storage and energy efficiency of mobile users’ devices.Through MEC’s computing offloading technology,mobile user devices can significantly reduce application delay,improve service quality,and meet the needs of mobile users.Through reasonable allocation of computing resources,each mobile user device can have sufficient MEC server computing resources and make efficient use of its own resources,so that users can enjoy high-quality service experience.The goal of this paper is how to design a reasonable scheme of computing uninstallation and resource allocation for mobile user devices when MEC server has limited computing resources.The main research content is computing uninstallation and resource allocation in two application scenarios:1.Aiming at the problem of how to choose the best computing offloading method for mobile user equipment and reasonably allocate computing resources of MEC server in single-cell scenario,a Q-learning algorithm based on reinforcement learning is proposed.Firstly,a communication model for the single-cell scenario is established.Secondly,a computing model is established for the mobile user equipment consisting of two computing execution modes: local computing and offload computing.Finally,a problem model with the optimization goal of minimizing the system weighted total cost is established.Through simulation experiments,under the condition of setting different system parameters,the results show that the system weighted total cost is smaller than that of the benchmark algorithm,and the performance is better than other algorithms,which verifies the effectiveness of the algorithm.2.Aiming at the ultra-dense network scenario,how to allocate the best MEC-small cell for mobile user equipment for offload computing and resource allocation,and the problem that the Q-learning algorithm may fall into the problem of dimensional disaster,a new scheme is designed to introduce the action classification algorithm into the DQN algorithm.First,establish a system model for ultra-dense network scenarios,and secondly establish a computing model for mobile user equipment consisting of local computing and offload computing.Different from the single MEC server in the single cell scenario,this scenario needs to be considered as a mobile user.The device allocates the best MEC-small cell from multiple MEC-small cells to minimize the optimization objective in the problem model,that is,the system weighted total cost.Through simulation experiments,under the condition of setting different system parameters,the results show that the proposed new scheme can obtain a smaller total system weighted cost than the Q-learning algorithm and the benchmark algorithm,which verifies the effectiveness of the algorithm.